The Federal Circuit delivered an important clarification on the enablement standard for prior art Agilent Technologies, Inc. v. Synthego Corp., No. 2023-2186 (Fed. Cir. June 11, 2025), affirming PTAB decisions that invalidated all claims of two CRISPR gene-editing patents. The case featured arguments by IP luminaries Mark Lemley and Edward Reines, with the patent challenger, Reines, coming up on top this time. The case distinguishes Amgen Inc. v. Sanofi, 598 U.S. 594 (2023): establishing clear boundaries between the enablement requirements for patent validity under 35 U.S.C. § 112 and the enablement standards for anticipatory prior art under § 102.
The unanimous Judge Prost decision also establishes that abandoned patent applications and research projects retain their full potency as prior art, rejecting arguments that abandonment should diminish their anticipatory effect. The patentee’s appellant brief had pointedly asked:
Did the Board err in finding that the claims were anticipated by and obvious over a prior art reference that never worked, was ultimately withdrawn, and which offered quadrillions of possible art combinations with no guidance to choose one that might work?
In rejecting the appeal, the court reinforced that prophetic examples in prior art can serve as anticipating references even in unpredictable fields, provided they contain sufficient enabling disclosure, and also concluded that the PTAB’s decision was based upon substantial evidence.
Background: Early CRISPR-Cas systems had a problem: guide RNA (gRNA) degradation. Agilent developed chemically modified gRNAs that resist degradation while maintaining functionality. The claims require both specific chemical modifications (phosphorothioate) and the functional capability to associate with Cas proteins and target specific DNA sequences. U.S. Patent Nos. 10,337,001 and 10,900,034
Synthego challenged these patents via IPR, relying primarily on a Pioneer Hi-Bred abandoned patent application (WO 2015/026885 A1) as anticipatory prior art. Pioneer Hi-Bred disclosed broad, prophetic examples of potential gRNA modifications but notably failed to demonstrate any working examples and was ultimately abandoned. According to Agilent, the 150 page reference did not simply teach the particular claimed modification but rather disclosed “over a quadrillion quadrillion possible combinations,” without guidance on which combinations would actually work.
Amgen Red Herring: Agilent’s primary appellate strategy involved invoking the Supreme Court’s recent decision in Amgen Inc. v. Sanofi, 598 U.S. 594 (2023), arguing that the Pioneer Hi-Bred reference failed to provide sufficient enablement because it disclosed numerous potential modifications without indicating which combinations would actually work. In Amgen, the Supreme Court invalidated functionally-defined antibody claims for lack of enablement, finding that the specification would have required “painstaking experimentation to see what works” across the entire claimed genus.
Judge Prost, writing for a unanimous panel, rejected this analogy. As she noted during oral argument:
I’m tired of Amgen. Amgen is different as a legal matter . . . in in terms of what you have to enable because in Amgen, maybe I’m wrong, but I recall, you had to look at the entire specification. It was a, it was a basic enablement case of the, a particular patent. This is different, and you’ve already agreed with me that you only have to enable in the prior art that specific portion that is relied on to establish anticipation.
The difference here – Section 112 enablement, at issue in Amgen, requires that a patent specification enable the full scope of the claimed invention to prevent patentees from obtaining broader exclusive rights than their disclosures justify. In contrast, anticipatory prior art need only enable “a single embodiment of the claim” to defeat novelty. Reaffirming In re Morsa, 803 F.3d 1374 (Fed. Cir. 2015). By rejecting Amgen‘s relevance, Agilent effectively establishes what might be called the “Amgen firewall”—a doctrinal barrier preventing the Supreme Court’s heightened enablement standards from migrating into prior art analysis.
Prophetic Examples in Abandoned Inventions are not Lesser Prior Art even in Unpredictable Fields
The Federal Circuit’s decision establishes a principle that might surprise many outside of patent law: abandoned patent applications containing primarily prophetic examples retain their full potency as prior art, even when those prophetic examples repeatedly prove incorrect and even in highly unpredictable technological fields. This “abandoned prophecy doctrine” has significant implications as we move deeper into the age of artificial intelligence because it opens the door to AI-powered prior art strategies.
To understand why this matters, we need to grasp what the court actually held about prophetic examples. Pioneer Hi-Bred’s patent application was essentially an elaborate research proposal—it described numerous theoretical gRNA modifications without demonstrating that any of them actually worked. When Pioneer Hi-Bred ultimately abandoned the application after failing to make their approach work, one might reasonably think this failure would diminish the application’s value as prior art. The Federal Circuit, applying traditional patent law doctrine, disagreed. But, the result here supports the AI prior art flood scenario.
As Mark Lemley noted during oral arguments, Pioneer Hi-Bred contained “prophetic examples throughout that say, ‘This would work, this would work, this would work.’ They turn out one after the other to be wrong or at the very least unproven.” Despite his explanation, the court held that this pattern of failed predictions did not diminish the application’s anticipatory effect. In this case, the prophetic examples appeared to stem from human researchers as part of an actual research endeavor, those are not doctrinal limiting principles. Rather, under Agilent, speculative disclosures retain full prior art potency even if they are ultimately proven to be fundamentally misconceived. All this seems like the court is putting on blinders and simply following its rule from in re Gleave that “[a] reference stands for all that it discloses to one of ordinary skill in the art.” 560 F.3d 1331 (Fed. Cir. 2009). Here, one example of a gRNA sequence covered by the claims is expressly disclosed in the prior art with a statement in the prior art that it would work if used. That was enough, at least in this case, for the appellate panel to find substantial evidence to affirm that the prior art enabled at least one embodiment of the claimed invention – a conclusion supported by the presumption-of-enablement.